Chapter 6
library(extrafont)
## Registering fonts with R
loadfonts()
## Charis SIL already registered with pdfFonts().
library(ggplot2)
library(reshape2)
library(grid)
library(dplyr)
##
## Attaching package: 'dplyr'
##
## The following objects are masked from 'package:stats':
##
## filter, lag
##
## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, union
Figure 1: Case study 1 mechanism —-
library(DiagrammeR)
nodes <- create_nodes(nodes = c("visibility", "authority", "legit. challenges", "transparency", "dem. state demands", "Secretariat"), shape = c("circle", "circle", "circle", "rectangle", "circle", "diamond"), fontname = "Charis SIL", fixedsize = T, fontsize = 4, x = c(1, 1, 2, 4, 3, 3), y = c(6, 4, 5, 5, 6, 5), fillcolor = c("White", "White", "White", "WhiteSmoke", "White" , "White"), style = "filled")
edges <- create_edges(edge_from = c("visibility", "authority", "legit. challenges", "dem. state demands", "Secretariat"), edge_to = c("legit. challenges", "legit. challenges", "Secretariat", "Secretariat", "transparency"))
graph <- create_graph(nodes_df = nodes, edges_df = edges, graph_attrs = c("layout = neato"))
render_graph(graph)
#graph.svg <- render_graph(graph, output = "SVG", width = 1800, height = 1600)
#write(graph.svg, file = "ch6-mechan1.svg")
Figure 2: IAEA in the Global News Media —-
hl_transparency <- read.csv2('../data/iaea_newspaper_headlines_searcht.csv')
plot.df <- melt(hl_transparency, id.vars = "Year")
p1 <- ggplot(plot.df) + geom_line(aes(x = Year, y = value, linetype = variable), size = 1) + theme_bw() + theme(legend.position = "bottom", legend.key.size = unit(0.5, "cm"), legend.key = element_blank()) + ylab("No. of Articles / No. of Hits") + xlab("Year") + theme(text=element_text(family="Charis SIL", size=12)) + scale_linetype_manual(name = "", values = c(1, 3), labels = c("IAEA headline articles", "Annual hits for transparency")) + annotate("text", x = 1986, y = 25, label = "Chernobyl", family="Charis SIL", size = 5, angle = 90) + annotate("text", x = 1994, y = 30, label = "PRK", family="Charis SIL", size = 5, angle = 90) + annotate("text", x = 2004, y = 90, label = "KOR, IRN", family="Charis SIL", size = 5, angle = 90) + annotate("text", x = 2003, y = 70, label = "IRQ", family="Charis SIL", size = 5, angle = 90) + annotate("text", x = 2007, y = 75, label = "PRK, IRN", family="Charis SIL", size = 5, angle = 90) + annotate("text", x = 2011, y = 93, label = "IRN, Fukushima", family="Charis SIL", size = 5, angle = 90) + coord_cartesian(ylim = c(0, 110))
p1
#ggsave(filename = "transp_headlines.pdf", plot = p1, width = 8, height = 6, dpi = 300)
Figure 3: IAEA transparency in GC debate —-
library(tm)
library(slam)
library(dplyr)
library(ggplot2)
library(reshape2)
library(grid)
# read in text files
corpus <- Corpus(DirSource("../data/corpora/iaea-gc-records/", encoding="UTF-8"), readerControl=list(language="en"))
# copy corpus and remove punctuation, numbers and transform to lower
dtmCorpus <- corpus
dtmCorpus <- tm_map(dtmCorpus, content_transformer(tolower))
dtmCorpus <- tm_map(dtmCorpus, content_transformer(function(x) gsub("(['’\n  ]|[[:punct:]]|[[:space:]]|[[:cntrl:]])+", " ", x)))
dtmCorpus <- tm_map(dtmCorpus, removeNumbers)
# create document term matrix and remove stop words
dtm <- DocumentTermMatrix(dtmCorpus, control=list(tolower=FALSE, wordLengths=c(2, Inf)))
dictionary <- data.frame(row.names=colnames(dtm), "Occurrences"=col_sums(dtm), "Stopword"=ifelse(colnames(dtm) %in% stopwords("en"), "Stopword", ""), stringsAsFactors=FALSE)
dtm <- dtm[, !colnames(dtm) %in% stopwords("en")]
attr(dtm, "dictionary") <- dictionary
meta(corpus, type="corpus", tag="language") <- attr(dtm, "language") <- "en"
meta(corpus, type="corpus", tag="processing") <- attr(dtm, "processing") <- c(lowercase=TRUE, punctuation=TRUE, digits=TRUE, stopwords=TRUE, stemming=FALSE, customStemming=FALSE, twitter=FALSE, removeHashtags=NA, removeNames=NA)
corpus
## <<VCorpus>>
## Metadata: corpus specific: 2, document level (indexed): 0
## Content: documents: 54
dtm
## <<DocumentTermMatrix (documents: 54, terms: 32937)>>
## Non-/sparse entries: 291895/1486703
## Sparsity : 84%
## Maximal term length: 74
## Weighting : term frequency (tf)
# create bi-gram "public information"
library(RWeka)
options(mc.cores=1)
BigramTokenizer <- function(x) NGramTokenizer(x, Weka_control(min = 2, max = 2)) # set function for bi-grams
bigrams <- DocumentTermMatrix(dtmCorpus, control = list(tokenize = BigramTokenizer)) # calculate bi-grams
inf_bigrams <- as.data.frame(as.matrix(bigrams))
inf_bigrams <- select(inf_bigrams, contains("information")) # filter for those including information
inf_bigrams$GROUP <- rep(1, 54)
# identify largest groups
groups <- inf_bigrams %>% group_by(GROUP) %>% summarise_each(funs(sum))
groups <- as.data.frame(t(groups))
summary(groups$V1)
## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 1.000 1.000 1.000 6.891 3.000 430.000
#dtm2 <- removeSparseTerms(dtm, 0.1) #alternatively, but not workable for this analysis
terms <- as.data.frame(as.matrix(dtm))
terms$Year <- 1958:2011
## combine relevant terms for transparency
terms$TRANSPARENCY <- terms$transparent + terms$transparence + terms$transparency + terms$transparently + terms$transparencyand
terms$DISCLOSURE <- terms$disclose + terms$disclosed + terms$disclosing + terms$disclosure + terms$disclosures
# add from information bigrams, selecting only relevant bigrams with a large number of hits
terms$PUBLIC.INFO <- inf_bigrams$`public information` + inf_bigrams$`information exchange` + inf_bigrams$`disseminate information` + inf_bigrams$`information center` + inf_bigrams$`provide information` + inf_bigrams$`information campaign`
#write.csv(terms, file = "terms.csv") # export full matrix, if needed.
terms2 <- terms %>% select(Year, TRANSPARENCY, DISCLOSURE, PUBLIC.INFO)
#write.csv(terms2, file = "~/owncloud/Diss/dissertation-book/ch6/terms2.csv")
library(extrafont)
p <- melt(terms2, id = "Year")
p1 <- ggplot(data = p, aes(x = Year, y = value, fill = variable)) + geom_bar(stat = "identity", position = "stack") + theme_bw() + theme(legend.position = "bottom", legend.key.size = unit(1, "cm"), legend.key = element_blank()) + ylab("Hits in annual GC records") + xlab("Year") + theme(text=element_text(family="Charis SIL", size=12)) + scale_fill_manual(values = c("#BDBDBD", "#353334", "#7E7C7D"), name = "")
p1
#ggsave(filename = "ch6-ms_transparency.pdf", plot = p1, width = 8, height = 6, dpi = 300)
Figure 5: IAEA statements vs. state demands —-
iaea.transparency.statements.export <- read.csv("../data/iaea-transparency-statements-export.csv", sep=";")
# overview for IAEA statements:
plot.df <- iaea.transparency.statements.export %>% group_by(Year, Code...Topic) %>% summarise(no.statements = n())
ggplot(plot.df) + geom_bar(aes(x = Year, y = no.statements, fill = Code...Topic), stat = "identity", position = "stack")
## combined plot
df1 <- state.transparency.statements.export %>% filter(Code...Topic == "ADM" | Code...Topic == "INSTR") %>% group_by(Year, Code...Topic, Democracy) %>% summarise(value = n())
df1$speaker[which(df1$Democracy == 1)] <- rep("Dem. States", length(df1$value[which(df1$Democracy == 1)]))
df1$speaker[which(df1$Democracy == 0)] <- rep("Non-Dem. States", length(df1$value[which(df1$Democracy == 0)]))
df2 <- iaea.transparency.statements.export %>% filter(Code...Topic == "ADM" | Code...Topic == "INSTR") %>% group_by(Year, Code...Topic) %>% summarise(value = n())
df2$speaker <- rep("IAEA", length(df2$value))
plot.df <- rbind_all(list(df1, df2))
## Warning in rbind_all(list(df1, df2)): Unequal factor levels: coercing to
## character
p1 <- ggplot(plot.df) + geom_bar(aes(x = Year, fill = speaker, y = value), size = 1, stat = "identity") + theme_bw() + theme(legend.position = "bottom", legend.key.size = unit(1, "cm"), legend.key = element_blank()) + ylab("No. of statements") + xlab("") + theme(text=element_text(family="Charis SIL", size=12)) + scale_fill_manual("Speaker", values = c("#BDBDBD", "#353334", "#7E7C7D")) + facet_wrap(~ Code...Topic )
p1
#ggsave(filename = "ch6-ms-iaea-transp-statements.pdf", plot = p1, width = 8, height = 6, dpi = 300)
Figure 7: mechanism case 2 —-
nodes <- create_nodes(nodes = c("inequality", "state demands", "Secretariat", "transparency", "o.g. norm"),
shape = c("circle", "circle", "diamond", "rectangle", "circle"),
fontname = "Charis SIL",
fixedsize = T,
fontsize = 4,
x = c(1, 2, 3, 4, 2),
y = c(4.5, 4.5, 5, 5, 5.5),
fillcolor = c("White", "White", "White", "WhiteSmoke", "White"),
style = "filled")
edges <- create_edges(edge_from = c("inequality", "state demands", "Secretariat", "o.g. norm"),
edge_to = c("state demands", "Secretariat", "transparency", "Secretariat"))
graph <- create_graph(nodes_df = nodes, edges_df = edges, graph_attrs = c("layout = neato"))
render_graph(graph)
#graph.svg <- render_graph(graph, output = "SVG", width = 1800, height = 1600)
#write(graph.svg, file = "ch6-mechan2.svg")
Figure 9: Third mechanism —-
nodes <- create_nodes(nodes = c("media attention", "politicization", "Secretariat", "GC part.", "res. constraints", "part. events"),
shape = c("circle", "circle", "diamond", "rectangle", "circle", "rectangle"),
fontname = "Charis SIL",
fixedsize = T,
fontsize = 4,
x = c(1, 2, 3, 4, 3, 4),
y = c(5.8, 5.8, 5.8, 5.5, 4.6, 4.9),
fillcolor = c("White", "White", "White", "WhiteSmoke", "White", "WhiteSmoke"),
style = "filled")
edges <- create_edges(edge_from = c("media attention", "politicization", "Secretariat", "res. constraints"),
edge_to = c("politicization", "Secretariat", "GC part.", "part. events"))
graph <- create_graph(nodes_df = nodes, edges_df = edges, graph_attrs = c("layout = neato"))
render_graph(graph)
#graph.svg <- render_graph(graph, output = "SVG", width = 1800, height = 1600)
#write(graph.svg, file = "ch6-mechan3.svg")
